WorldTune

Internet radio streaming technology.

Pressing play on an online station feels like nothing at all. Behind that second there’s a studio, an encoder, a streaming server, a codec, a little text about the current song, and a player that has to understand all of it. Here’s the tour, with as little jargon as we could manage.

Illustration of the streaming chain: studio, encoder, server, and listeners

The chain, start to finish

Almost every station follows the same four steps. Something makes sound: a studio desk, a DJ’s laptop, or a piece of automation software playing a schedule. An encoder squeezes that sound into a stream. A server hands that stream out to everyone who connects. And your player grabs a few seconds of it, buffers, and starts playing.

The parts vary. A community station might run Icecast on a cheap server with a free encoder. A big network might pay a streaming provider to do the whole thing. A station that cares most about phones might chop the audio into HLS segments so it can ride on ordinary web infrastructure. Same chain, different parts.

Icecast, the open-source workhorse

Icecast is the server a huge share of independent stations run on. It takes the encoded audio from the studio and copies it out to every listener who connects. One Icecast server can carry several streams at once, each on its own “mount point”, which is why you sometimes see the same station offered at two bitrates or in two formats.

It’s free, it runs on a modest Linux box, and nearly every encoder knows how to talk to it. That’s why you’ll find it behind college stations, community broadcasters, and plenty of hobby projects. It handles plain MP3 streams just as happily as newer open formats like Opus.

Diagram of an Icecast server with three mount points serving listeners
One source in, many listeners out. That’s the whole job of a streaming server.
Diagram of an encoder feeding a SHOUTcast DNAS server, with listeners and a public directory listing
SHOUTcast gave small stations two things at once: a server and a place to be found.

SHOUTcast, the one that started it

SHOUTcast came out of the Winamp world in 1998 and, for a lot of people, was the first time internet radio felt real. The server software receives a stream from an encoder, serves it to listeners, and can report the station and the current song to a public directory so people can find it.

A surprising number of stations still run it, or run something that speaks its protocol, because it works and because almost every player on earth understands it. For you as a listener it means a plain stream link, a song title when the station bothers to send one, and no compatibility surprises. We wrote a whole guide about SHOUTcast if you want the long version.

HLS and DASH: audio in slices

HTTP Live Streaming, or HLS, takes a different approach. Instead of one endless stream, the audio is cut into short files, a few seconds each, listed in a playlist that keeps updating. Any ordinary web server or CDN can serve those files, which is exactly why large broadcasters like it: it scales like a website, and it copes well with flaky mobile connections.

MPEG-DASH does the same thing as an open international standard. You’ll meet it more often in video than in radio, but the idea is identical: small segments, fetched one after another, with the player free to switch to a lighter version if your connection gets worse.

Diagram of an HLS playlist, numbered audio segments, and a bitrate ladder
Segmented streaming trades a little delay for the ability to scale like any other website.

Encoders, automation, and codecs

The server only hands out what it’s given, so a lot of the real work happens before it. Tools like Liquidsoap run the schedule, crossfade the songs, insert jingles, and fall back to a playlist if the live feed drops. FFmpeg shows up everywhere too, converting, inspecting, and repackaging audio for whatever the station needs to send.

The codec is how the audio gets squeezed. MP3 is old but plays on absolutely everything. AAC sounds better at the same bitrate and is common on commercial stations. Opus is the newer, open, royalty-free option that does remarkably well at low bitrates, which makes it popular with talk stations and anyone watching their bandwidth bill.

The text that rides along with the sound

A stream can carry more than audio. Many stations tuck a line of text into it every few seconds, usually the artist and title, sometimes the show name. Players read that line, and apps like WorldTune use it to look up cover art, artist details, and lyrics from outside sources.

That’s why one station shows a full now playing panel and the next shows only its name. The second station simply isn’t sending the text, and there’s nothing a player can do about that. WorldTune shows what each broadcaster sends and leaves the rest alone.

Diagram comparing the text fields a station sends with the now playing panel a listener sees
Artist, title, artwork, and lyrics all trace back to a few words the station sends with the audio.

What you’ll notice as a listener

Some stations start faster than others

A plain MP3 or AAC stream usually starts in a second or two. A segmented HLS stream has to fetch a playlist and a couple of files first, so it can take a little longer and run a little further behind live.

The song title isn’t guaranteed

If a station doesn’t send the track name, no player can show it. If it sends it late or wrong, you’ll see it late or wrong. It’s the broadcaster’s setup, not yours.

Old formats stick around for a reason

Stations pick what works with the most players, costs the least to host, and what their one technical volunteer knows how to run. That’s why MP3 over Icecast or SHOUTcast is still everywhere.

To hear all of this in action, open the web player and watch the now playing line change between stations. The listening guide covers the practical side, and the history of online radio covers how we got here.